Study On Spatiotemporal Changes Of Bamboo Forest In Jiangxi Province Based On Remote Sensing

Bamboo forest is an important forest resource.It has the characteristics of rapid growth and Carbon sequestration effect.Obtaining bamboo forest resources and its spatial distribution are of great significance for the distribution of bamboo product processing industry and evaluation of bamboo forest ecosystem service functions.This paper takes the bamboo forest in Jiangxi Province as the research object,and carries out the following research work under the support of the GEE cloud platform:(1)Using the images acquired by Landsat8-OLI sensor in 2016,we analyzed the seasonal variation characteristics of spectral characteristics and spectral vegetation index of bamboo forest and the differences of other features,and explored the feasibility of extracting bamboo forest information from remote sensing images;(2)Using Landsat remote sensing images,comparing and analyzing the minimum distance method and random forest algorithm to extract the accuracy difference of bamboo forest in Jiangxi Province in 2016,the effects of different classification characteristics on the extraction accuracy of bamboo forests were also compared and analyzed.(3)Using the modified AASG algorithm to obtain bamboo forest samples from 1995and 2005,the spatial distribution of bamboo forest was extracted by random forest algorithm,and analyzed the distribution characteristics of bamboo forest from 1995 to2016.Research indicates:(1)The bamboo forest and other land types are separable,and the ideal bands are the4th,5th,and 6th bands of OLI data;the spectral index NDWI and NDVI can increase the difference between bamboo forest and other land features,especially the value of NDWI and the value of NDVI in autumn and winter.Contrast and Variance in texture features help distinguish between bamboo and other forest types.When remote sensing images are used to extract bamboo forest information,the remote sensing images in winter are the best and can be divided.(2)Compared with the minimum distance method,the precision of bamboo forest extraction by random forest algorithm is improved by 20%,and the advantages are obvious.Compared with single-phase remote sensing image,using the minimum cloud image synthesis can effectively improve bamboo forest classification accuracy and use random forest.Using random forest algorithms and dense Landsat time series images,and adding NDVI,NDWI,DEM,and texture features as auxiliary input features have the best effect on extracting the spatial distribution of bamboo forests.(3)The bamboo forest distribution maps of 1995 and 2005 extracted using Random Forest Algorithm and dense Landsat time series imagery remote sensing,the overall accuracy was 0.82 and 0.84 respectively,and the KAPPA coefficients were 0.69 and0.75,respectively.The results show that the AASG algorithm can be used to obtain the stable features in the historical period.(4)Since 1995,bamboo forest in Jiangxi Province has been rapidly expanding.In1995,the bamboo forest area was 3,769.19km~2.In 2005,the bamboo forest area expanded to 4901.40km~2.In 2016,the bamboo forest area expanded further to9,988.52km~2,with an average annual increase of 295.30km~2.The average growth rate is 7.8%,exceeding the global bamboo growth rate.(5)There are bamboo forests distributed in most counties and cities of Jiangxi Province,and bamboo forests are widely distributed.The counties with more bamboo forests include:Yanshan County,Tonggu County,Pingxiang,Wuning County,Zixi County,and Chongyi County;The elevation range of the distribution is 300-900m,accounting for 77.26%of the total area of the bamboo forest.From the perspective of slope and slope,the bamboo forests are mostly distributed on slopes(slopes of16°-25°)and on steep slopes(slopes of 26°-35°),accounting for 70.19%of the total area of the bamboo forest,and is more like that in the north,southeast,and northwest.